Credit, Bankruptcy, and Intermediary Market...
Transcript of Credit, Bankruptcy, and Intermediary Market...
Credit, Bankruptcy, and Intermediary Market Structure
Andrea Buraschi
London Business School and CEPR
Li, Hao
University of Toronto
June, 2004
Abstract: We use a model of costly monitoring to study the determinants of savings
mobilization, capital allocation and entrepreneur bankruptcy rates under different mar-
ket structures of financial intermediation. Borrowers have access to the same investment
project but differ in the value of their collateralizable assets. The main finding is that
monopolistic intermediation induces less lending and higher bankruptcy rates than com-
petitive intermediation. These two types of monopoly distortions are due to the monopoly
power with lenders and with borrowers respectively. Under both market structures, an
increase in available credit or a reduction in monitoring costs implies that more borrowers
obtain funds. Bankruptcy rates, however, are reduced only under competitive intermedi-
ation.
JEL classification: G2, D3, D4, P26, P50
Keywords: Credit market structure, banking, bankruptcy, default, collateral
Acknowledgements: Andrea Buraschi is an Assistant Professor from London Busi-
ness School, Institute of Finance, Sussex Place, Regent’s Park, London NW1 4SA, United
Kingdom; Li, Hao is a Professor of Economics at the University of Toronto. We grate-
fully acknowledge the suggestions of In-Koo Cho, Bruce Greenwald, Lars Hansen, Robert
Lucas, George Pennacchi, Robert Townsend, and Michael Woodford. The Editor and an
anonymous referee made detailed comments that have helped us improve on an earlier
draft.
– i –
Credit, Bankruptcy, and Intermediary Market Structure
Abstract: We use a model of costly monitoring to study the determinants of savings
mobilization, capital allocation and entrepreneur bankruptcy rates under different mar-
ket structures of financial intermediation. Borrowers have access to the same investment
project but differ in the value of their collateralizable assets. The main finding is that
monopolistic intermediation induces less lending and higher bankruptcy rates than com-
petitive intermediation. These two types of monopoly distortions are due to the monopoly
power with lenders and with borrowers respectively. Under both market structures, an
increase in available credit or a reduction in monitoring costs implies that more borrowers
obtain funds. Bankruptcy rates, however, are reduced only under competitive intermedi-
ation.
JEL classification: G2, D3, D4, P26, P50
Keywords: Credit market structure, banking, bankruptcy, default, collateral
– ii –
1. Introduction
In an economy with a well-developed financial intermediary sector, lenders rarely lend to
borrowers directly. Instead, as financial intermediaries between lenders and borrowers,
banks perform the important function of insuring lenders against the risks of default by
borrowers. This function is achieved by screening the investment projects of borrowers, by
assessing their credit worthiness and by monitoring the investment outcomes. How is the
function affected by the market structure of the intermediary sector?
A recent line of literature on financial intermediation has stressed the effect of banking
competition on project screening and found that competition in the intermediary sector
increases default risks. For example, in Broeker’s (1990) model, banks announce interest
rates at which they will provide credit to borrowers that pass a credit test. When the
credit tests are independent across banks and give noisy signals of the credit worthiness of
the borrowers, the proportion of borrowers that pass at least one test increases with the
number of banks. Thus, when there is a greater number of competing banks, the average
credit quality decreases, and hence the default risks increase. Riordan (1993) considers a
similar setting and shows that in the absence of collateral or other self-selection devices,
banks have the incentive to cut interest rates to improve the credit quality of the pool of
their borrowers. The intuition is that borrowers that accept the least favorable interest
rate are the ones that have been rejected by other banks. In this setting, more competition
makes each individual bank more conservative about marginal loans. This “winner’s curse”
becomes worse with more competing banks, and the aggregate result is less bank lending
(see also Shaffer, 1998 and Marquez, 2002).
In this paper we present a contrasting view on the effect of competition on the banks’
function of insuring lenders against risks of default by borrowers. Our paper differs from
the existing literature by focusing on the role of collateral in credit assessment and project
monitoring, while abstracting from project screening by assuming that borrowers have
identical investment projects. We build on the costly monitoring framework introduced
by Townsend (1979).1 In Townsend’s (1979) original contribution, borrower-entrepreneurs
1 Williamson (1987b) uses a similar costly monitoring framework in a real business cycle setup to show
– 1 –
have ex post private information about the outcome of their projects, and monitoring (au-
diting) is costly for lenders. Diamond (1984) show that intermediaries arise as “delegated”
monitors to economize on the monitoring cost and solve the free-rider problem among
lenders; we use this result to compare alternative market structures of the intermediary
sector. To the costly monitoring framework, we introduce the feature that entrepreneurs
differ in the amount of non-liquid wealth that they can lodge as collateral in loan contracts.
In our model, entrepreneurs are endowed with fixed assets that are needed for production
and therefore cannot be liquidated to finance the project, but they can use their fixed assets
as collateral to reduce the cost of the ex post moral hazard problem to the intermediary
in monitoring the project output.
Our main discovery is that monopolistic intermediation results in less lending and
higher entrepreneur bankruptcy rates than competitive intermediation. A monopoly in-
termediary lends less because it has the monopoly power with lenders and perceives a
higher marginal cost of credit than competitive intermediaries. In contrast, the monopoly
distortion of higher business bankruptcy rates arises from the market power with borrow-
ers. As the monopoly intermediary extracts more surplus from entrepreneurs by raising
debt repayments, entrepreneurs are more likely to forfeit collateral and declare default. In
other words, the monopoly intermediary has incentives to demand a high debt repayment
from a borrower so as to sometimes induce default and seize the collateral, while under
competitive intermediation intermediaries compete to lower the debt repayment and thus
reduce the default risks. The monopoly distortion of higher bankruptcy rates exists even
if the monopoly intermediary faces a competitive credit market and has no market power
with lenders. Further, in a market expansion due to either an increase in the credit supply
or a decrease in the monitoring cost, the default rate for the pre-existing loans is unaffected
under monopolistic intermediation while it falls under competitive intermediation.
Our result that monopolistic intermediation increases bankruptcy risks is related to
Rajan (1992). He considers a model in which a bank obtains monopolistic information
that endogenous financial intermediation helps to account for monetary phenomena such as the positivecorrelation between the price level and real output. Bernake and Gertler (1989) address similar questionswith a model of collateral.
– 2 –
on the borrower which can be used to gain control on the borrower’s decisions and in-
duce the borrower to terminate negative NPV projects. Unfortunately, this control also
allows the bank to extract surplus from the borrower ex post. This makes the borrower
exert less than optimal effort, thus reducing project returns. Since competition among
banks reduces their bargaining power over the borrower, Rajan’s model has the implica-
tion that competitive intermediation can lead to greater effort in project selection and
hence lower bankruptcy rates. In the present paper there is no project selection effort,
and the monopoly intermediary’s tendency of inducing higher bankruptcy rates does not
arise from a holdup problem. Instead, we have a model in which collateral is used to
alleviate the ex post project monitoring problem, and the monopoly distortion of higher
bankruptcy rates comes from the monopoly intermediary’s incentives to demand a high
debt repayment so as to seize the collateral in the event of default. A further difference
between our model and Rajan’s is that in our paper the optimal monopoly repayment
schedule and the competitive repayment schedule are determined in a general equilibrium
framework with both the lenders and the borrowers explicitly modeled.
The model is presented in the next section. Sections 3 and 4 analyze monopolistic
intermediation and competitive intermediation respectively. We characterize the loan con-
tracts and deposit interest rate for each intermediary structure. Section 5 compares the
two representative structures in terms of aggregate lending and entrepreneur bankruptcy
rates, and discusses how changes in available credit and monitoring cost affect credit mar-
ket performance under the two structures. Two extensions of the main results are briefly
discussed. In Section 6 we discuss the related extensive literature on “excessive” compe-
tition in the financial intermediary sector, and provide some evidence for our result that
competition increases bank lending. We conclude with some remarks on the directions of
possible future work.
2. A Costly Monitoring Model with Collateral
Our economy is described by three types of agents (borrowers, intermediaries and lenders),
three types of goods (investment good, collateral and consumption good) and three periods
– 3 –
(contract period, production period and consumption period). In the contract period, bor-
rowers offer collateral to intermediaries to obtain the investment good, and intermediaries
in turn receive the investment good from lenders by promising repayment with interest.
In the production period, borrowers who received the investment good implement a risky
project and observe the output. In the consumption period, consumption takes place after
the contracts are settled between the intermediaries and the borrowers, and between the
intermediaries and the lenders. The details of these activities and the parameters of the
economy are described next.
There is a continuum of risk-neutral borrower-entrepreneurs, each of whom has access
to a risky project. The project takes one unit of the investment good and transforms it
into a non-negative random output of w units of the consumption good. The output w
is assumed to be i.i.d. across projects. The range of distribution of w is [w1, w2], with
mean w. The distribution function is F (w); we assume that the the corresponding density
function f(w) is continuous. Each entrepreneur can observe costlessly the outcome of
his project, but output verification costs g > 0 units of the consumption good for any
other agent. To focus on the role of collateral in capital allocation, we assume that this
monitoring cost is the same for all projects. The monitoring cost plays an important role
in our model, and should be interpreted to include not only the cost of auditing, but also
the transaction cost of bankruptcy to the intermediary.2
Entrepreneurs rely entirely on outside funding to provide one unit of the investment
good that is needed to carry out the project. However, they are endowed with different
units of divisible and non-liquid fixed assets that can be (partially or wholly) used as
collateral in the loan contract with lenders. These assets are non-liquid because they
are necessary for the production and cannot be sold to finance the project.3 Examples
2 Bankruptcy costs can be a substantial portion of the claims. Altman (1984) finds that the averageadministrative bankruptcy costs are 6% of the claims under Chapter 11 and 6.2% under U.K. Receivership.Franks and Torous (1994) discuss the direct and efficiency bankruptcy costs under alternative bankruptcycodes. They find that U.S. companies spend an average of 27 months in Chapter 11 and 17 months inworkouts outside the bankruptcy process, while the U.K. Insolvency Code generally results in a speediersettlement of claims.
3 Since all entrepreneurs have access to the same investment project, the role of collateral in ourmodel is different from that in Besanko and Thakor (1987), where borrowers with private informationregarding the prospects of their investment projects use collateral as a signal of credit-worthiness. Our
– 4 –
of such assets are patents or lab equipment of a biotech firm, or airplanes of an airline
company. For analytical convenience, we assume that the non-liquid assets are liquidated
after completion of the project, and the consumption value of the liquidated collateral can
be costlessly verified by lenders.4 Collateral represents the lender’s claim on the return
of the borrower’s project, when the latter is unable to honor the repayment obligation
specified in the loan contract. The range of the liquidation value of collateralizable assets
is [k1, k2], in units of the consumption good. The measure of entrepreneurs with non-liquid
wealth greater than k, denoted as U(k), is a decreasing continuous function of k. As we will
see, putting up more non-liquid assets as collateral helps alleviate the ex post monitoring
problem and lowers the cost of capital to entrepreneurs.
Motivated by the delegated monitoring result of Diamond (1984), we assume that
there is no direct contact between lenders and entrepreneurs; instead, the supply of credit
meets with the demand for loans indirectly through intermediaries, who offer contracts to
both borrowers and lenders. As they deal with a continuum of borrowers with independent
projects, the intermediaries in our model can perfectly diversify and offer lenders a risk-free
return.5 The contracts between intermediaries and lenders are therefore characterized by a
risk-free deposit arrangement. We represent lenders by a non-decreasing and differentiable
aggregate savings function S(r), where r is the risk-free deposit rate. The savings S(r)
are measured in the investment good. Since each entrepreneur needs one unit of the
investment good, the supply of credit represented by the function S(·) and the demand for
loans represented by the function U(·) are measured in the same unit. The savings function
S(r) can be substantiated by specifying endowments and preference for a representative
consumer. We simply take S(r) as a primitive, and assume that it is defined for all non-
negative deposit rates. We also make the following technical assumption:
model also differs from the ex ante moral hazard model of Holmstrom and Tirole (1997), who assume thatentrepreneurs’ own funds are entirely liquid and can be used to finance the project together with outsidefunds. In their model, the more own funds entrepreneurs have, the more credible their commitment of notmisusing the outside fund, and the more likely they can obtain outside funds.
4 The same assumptions on collateral are made, for example, in Schmidt-Mohr (1997). Adding adisparity in collateral valuation by the entrepreneur and by the lender, as in Barro (1976), will not changethe results in the present paper. This will be discussed in Section 5.3.
5 For results on intermediation when diversification is imperfect, see Winton (1997) and Yosha (1997).
– 5 –
Assumption 2.1: r + S(r)/S′(r) increases with r.
Note that the total cost of obtaining loans from obtaining loans from depositors at deposit
rate r is rS(r). Since the aggregation savings function S(r) is non-decreasing, Assumption
2.1 implies that the marginal cost of loans is increasing in the deposit rate. As can be
seen in the next section, this is sufficient for the second-order condition of the monopoly’s
optimization problem. Assumption 2.1 is satisfied as long as the slope of the savings
function decreases in the deposit rate, or does not increase too fast.
2.1. Risky debt
The building block for the subsequent sections is the loan contract between an intermediary
and an entrepreneur. Regardless of the collateral value k put up by the entrepreneur, any
optimal contract that maximizes a weighted sum of the expected profits of the entrepreneur
and the intermediary subject to the participation constraints of the two takes the form
of “risky debt.”6 A risky debt contract is characterized by a single number x, called the
“face value of debt.” If the realized output w exceeds x − k, the entrepreneur declares
“success” and pays x. If w < x− k, the entrepreneur “defaults” or declares “bankruptcy.”
In this case, the intermediary incurs the cost g to verify the output and gets all the output
plus collateral k. The logic behind the risky debt contracts is that it is in the interest of
the entrepreneur as well as the intermediary to discourage the entrepreneur from falsely
declaring bankruptcy. To do so requires making the punishment for default as severe as
possible, which is to forfeit all collateral and any output from the project. The proof
that any optimal contract takes the form of risky debt is given by, for example, Townsend
(1979), Gale and Hellwig (1985), Williamson (1987a, 1987b). The adaptation of the proof
to our model of intermediation with collateral is straightforward and omitted.
Let P (x, k) be the expected payment from an entrepreneur with collateral value k to
the intermediary under a risky debt contract with face value x, before the intermediary
6 We rule out stochastic monitoring for technical reasons. As Townsend (1979) first pointed out,random monitoring can be more efficient. We note that the focus here is comparison of alternative struc-tures of intermediation, and that any efficiency gain will occur under both monopolistic and competitiveintermediation, although the gain may differ.
– 6 –
incurs the monitoring expenses. Then,
P (x, k) =
x, for x ≤ w1 + k∫ x−k
w1(w + k)f(w) dw +
∫ w2
x−kxf(w) dw, for x ∈ (w1 + k, w2 + k)
w + k, for x ≥ w2 + k.
(2.1)
The probability of default is 0 if x ≤ w1+k, strictly between 0 and 1 if x ∈ (w1+k, w2+k),
and 1 if x ≥ w2 + k. A higher face value x leads to a greater probability of default for
x ∈ (w1 + k, w2 + k), and a greater expected payment P for x < w2 + k. Denote the
expected profits of the entrepreneur with collateral k from the risky debt contract with
face value x as B(x, k). Then, B(x, k) is the difference between the expected output w
and P (x, k):
B(x, k) = w − P (x, k). (2.2)
Since P (x, k) weakly increases with x, B(x, k) weakly decreases with x. Denote the gross
return to an intermediary of a risky debt with face value x and collateral k as L(x, k).
Then, since the intermediary incurs the cost of auditing the output, L(x, k) is the difference
between the expected payment P (x, k) and the expected monitoring expense:
L(x, k) = P (x, k)− gF (x− k). (2.3)
Finally, let T (x, k) = B(x, k) + L(x, k) be the sum of the entrepreneur’s profits and the
expected loan return, given by
T (x, k) = w − gF (x− k). (2.4)
Note that the total gross profit T decreases with x for x ∈ (w1+k, w2+k) because a higher
face value means a greater default probability and hence a greater expected monitoring
expense.
For x ∈ (w1 + k, w2 + k), we can rewrite L(x, k) as:
L(x, k) =∫ x−k
w1
(v − g)f(v) dv +∫ w2
x−k
(x− k)f(v) dv + k,
Note that the first two terms depend on face value x only through x − k. Thus, x ∈(w1 + k, w2 + k) maximizes L(x, k) for a given k if and only if x− k ∈ (w1, w2) maximizes
– 7 –
the first two terms above. The following assumption implies that for any collateral value
k, the gross return L(x, k) is a concave function of face value x.
Assumption 2.2:∫ w
w1(v − g)f(v) dv +
∫ w2
wwf(v) dv is strictly concave in w on [w1, w2]
and achieves maximum at a unique interior point w∗.
In our costly monitoring model, a higher face value of debt increases the probability that
the entrepreneur defaults, and can therefore decrease the loan return to the intermediary
due to greater monitoring expenses. Assumption 2.2 is a simple way of capturing this
idea. Although global concavity of the loan return is a strong assumption, it simplifies
the analysis of loan contracts and allows us to focus on the comparison of monopolistic
versus competitive intermediation.7 Note that in our model all entrepreneurs have the
same investment project, which explains why Assumption 2.2 is stated independently of
the collateral value. The assumption that the loan return function is concave has been a
standard one since Stiglitz and Weiss (1981).8
By taking derivatives of the expression in Assumption 2.2 with respect to w, we find
that w∗ satisfies
1− F (w∗)− gf(w∗) = 0. (2.5)
Regardless of the intermediary market structure, for any collateral value k, the maximum
face value of debt that will occur in a loan contract is w∗ + k: if the face value of debt
exceeds w∗ + k, then since by Assumption 2.2 L is concave in x with maximum reached
at x = w∗ + k and since B decreases with x, a reduction in x would increase profits for
7 Regardless of the cost of monitoring g, concavity is assured as long as the density f(·) is non-decreasing, such as the uniform distribution. If f(·) is decreasing, concavity still obtains if it does notdecrease too fast or g is not too great. The function in Assumption 2.2 is always decreasing around w2.For it to have an interior maximizer on (w1, w2), the function has to be increasing around w1. A sufficientcondition is gf(w1) < 1.
8 Stiglitz and Weiss (1981) argue that credit rationing may occur when the market interest rate doesnot equate the supply of funds to the demand. They discuss two reasons: adverse selection of borrowers,where a higher interest rate attracts borrowers who are on average less likely to repay the loan, and moralhazard of borrowers, where a higher interest motivates borrowers to select riskier projects. In both cases,an increase in the interest rate will decrease the return to the intermediary from loans, and may thereforecontradict the zero-profit condition for intermediaries. Here, as in Williamson (1987a, 1987b) and Bernakeand Gertler (1989), costly output verification plays a similar role as adverse selection and moral hazard ofborrowers.
– 8 –
both the intermediary and the entrepreneur. It follows that the maximum default rate in
any loan contract is F (w∗). For this reason, we call the loan contract with the face value
w∗ + k the “maximum default contract.” The maximum default contract plays a special
role in our analysis. Define
p∗ =∫ w∗
w1
wf(w) dw +∫ w2
w∗w∗f(w) dw.
Then, under the maximum default contract, we have:
P (w∗ + k, k) = p∗ + k;
L(w∗ + k, k) = p∗ + k − gF (w∗);
B(w∗ + k, k) = w − (p∗ + k).
We can verify that p∗ lies between w1 and w.9 For future references, we denote
k∗ = w − p∗;
t∗ = w − gF (w∗).
Then, k∗ is the highest collateral value such that the maximum default contract gives
the entrepreneur non-negative expected profits, and t∗ = T (w∗ + k, k) is the sum of the
expected profits of the entrepreneur and the loan return to the intermediary under the
maximum default contract.10
3. Credit Market with Monopolistic Intermediation
A monopoly intermediary maximizes its expected profits by choosing a deposit rate to
obtain savings from lenders and by offering a menu of contracts to entrepreneurs through
9 By taking the derivatives with respect to w∗ of the equation that defines p∗, we can show that p∗is an increasing function of w∗. Further, w∗ = w1 implies that p∗ = w1, and w∗ = w2 implies p∗ = w.Since w∗ is in (w1, w2), p∗ is between w1 and w.
10 We can show that t∗ > 0 so that gains of trade exist between depositors and the entrepreneur(through intermediaries) under the maximum default contract, regardless of the collateral value k ofthe entrepreneur. Indeed, we have t∗ > k∗ + w1, because by Assumption 2.2 at w = w∗ the function∫ w
w1(v − g)f(v) dv +
∫ w2
wwf(v) dv reaches the maximum of p∗ − gF (w∗), which is greater than w1, the
value of the function at w = w1.
– 9 –
which to lend the capital. The monopoly intermediary makes take-it-or-leave-it offers to
entrepreneurs according to their collateral value. This assumes that the value of collater-
alizable assets of each entrepreneur is known to the intermediary.11 We first characterize
the monopoly loan contracts and then the deposit rate.
3.1. Monopoly loan contracts
For an entrepreneur with any collateral k, the monopoly intermediary chooses the face
value of debt x to maximize the expected loan return L(x, k) subject to the entrepreneur’s
participation constraint that his expected profit B(x, k) is non-negative. The loan return
L as a function of x is given by (2.3). It is convenient to rewrite the entrepreneur’s
participation constraint as L(x, k) ≤ T (x, k), where T , given by (2.4), is the sum of the
loan return L to the intermediary and the expected profits B of the entrepreneur.
Figure 1 plots L(x, k) and T (x, k) for a given k, as functions of face value x, for
the case in which the random output w is uniformly distributed on [w1, w2]. How the
two functions intersect with each other depends on k, as we will see in the proof of the
following proposition. We can verify that the uniform distribution satisfies Assumption
2.2 if g < w2 − w1. In this case, from equation (2.5) we have w∗ = w2 − g, which is
between w1 and w2, as desired. Under the uniform assumption, for face value of debt
x ∈ [w1 + k, w2 + k], the total gross profit T is linear and the loan return L is quadratic.
These properties are special, but for any output distribution F that satisfies Assumption
2.2, T is decreasing and L is concave with the maximum reached at x = w∗ + k.
Let xm(k) denote the solution to the problem of maxx L(x, k) subject to L(x, k) ≤T (x, k), as a function of collateral value k. We now show that xm(k) is given by
xm(k)
= w∗ + k, for k < k∗
satisfies B(xm(k), k) = 0, for k ∈ [k∗, w − w1)
= w, for k ≥ w − w1.
(3.1)
The following proposition gives this result in words.
11 Our comparison results are robust to the alternative assumption that the collateral value is privateinformation and loan contracts must give entrepreneurs incentives to reveal it. See Section 5.3.
– 10 –
x
T
L
w1+k w*+k w2+k
Figure 1. L(x, k) and T (x, k) as functions of x for fixed k.
Proposition 3.1. Under monopolistic intermediation, among entrepreneurs that obtain
loans, (i) maximum default contracts are optimal for low collateral values, (ii) expected
profits are zero for intermediate collateral values, and (iii) the probability of default is zero
for high collateral values.
Proof. Consider the two functions L(x, k) and T (x, k) for any given k, as x varies from
0 to above w2 + k. Since the two functions are continuous, and since L(0, k) = 0 < w =
T (0, k) and L(w2 + k, k) = w − g + k > w − g = T (w2 + k, k), there is at least one
intersection in (0, w2 + k). There cannot be two or more intersections. To see this, note
that since L(x, k) is increasing and T (x, k) is decreasing in (0, w∗+k], there can be at most
one intersection in this range. If there is one, then there cannot be another intersection in
(w∗ + k, w2 + k), because both L(x, k) and T (x, k) are decreasing in [w∗ + k,w2 + k) but
T (x, k) decreases faster (compare (2.3) with (2.4); note that P (x, k) is increasing in x.)
For the same reasons, if there is no intersection in (0, w∗ + k] (in which case T is entirely
above L in this range), then there is exactly one intersection in (w∗ + k, w2 + k).
The two functions L(x, k) and T (x, k) can intersect in (0, w1 + k], (w1 + k, w∗+ k), or
[w∗ + k,w2 + k), depending on the collateral value k. The solution xm(k) to the problem
of maxx L(x, k) subject to L(x, k) ≤ T (x, k) depends on where the intersection is and thus
also on k. For k ≥ w − w1, we have L(0, k) < T (0, k) and L(w1 + k, k) = w1 + k ≥ w =
T (w1 +k, k), so the intersection occurs in (0, w1 +k]. Since L(x, k) = x and T (x, k) = w in
– 11 –
this range, the intersection is at x = w. In this case, L increases in x for x < w and L > T
for x > w, so we have xm(k) = w. For k ∈ [k∗, w−w1), we have L(w1+k, k) < T (w1+k, k)
and L(w∗ + k, k) = p∗ + k − gF (w∗) ≥ t∗ = T (w∗ + k, k), so the intersection occurs in
(w1 + k,w∗ + k]. In this case, L is increasing in x before the intersection and L > T
after the intersection, so the intersection gives xm(k). It satisfies B(xm(k), k) = 0 because
L = T at the intersection. Finally, for k < k∗, we have L(w∗ + k, k) < T (w∗ + k, k) and
L(w2 + k, k) > T (w2 + k, k), so the intersection occurs in (w∗ + k, w2 + k). Since in this
range L is decreasing in x, the solution xm(k) is not given by the intersection. Instead, it
is given by w∗ + k, because this maximizes L while satisfying L ≤ T . Q.E.D.
Proposition 3.1 can be explained as follows. In lending to the entrepreneurs, the
monopoly intermediary uses the face value of debt in the loan contract to discriminate
among different collateral values. Under Assumption 2.2, to maximize the return L on
a loan to an entrepreneur with collateral k, the monopolist should choose the maximum
default contract with face value w∗ + k if it satisfies the entrepreneur’s participation con-
straint. This is precisely the case for entrepreneurs with little collateral (k < k∗)—recall
that k∗ is the highest collateral value such that the maximum default contract yields
non-negative expected profits to the entrepreneur. The participation constraint of these
entrepreneurs does not bind under the monopoly loan contract. For all entrepreneurs
with collateral values higher than k∗, the maximum default contract with x = w∗ + k
would leave them with negative profits. In this case, the monopolist optimally lowers
the face value of debt from w∗ + k to the point xm(k) where the participation constraint
just binds, i.e., B(xm(k), k) = 0. For entrepreneurs with intermediate values of collateral
(k∗ ≤ k < w − w1), the participation constraint binds at a point where the probabil-
ity of default is strictly between 0 and 1. (This is the case depicted in Figure 1.) For
those entrepreneurs who can provide sufficient collateral (k ≥ w − w1), the participation
constraint binds at a point where “full collateralization” is achieved: the intermediary
extracts all output from the entrepreneur by charging w, and the collateral committed by
the entrepreneur is so high that he will never default.
Under the optimal schedule of face value characterized above, the loan return to the
monopoly intermediary can be shown to increase with collateral value k. For collateral
– 12 –
values below k∗, the intermediary obtains the highest return k + t∗ − k∗ on the loan to an
entrepreneur with collateral k, under the maximum default contract with face value w∗+k.
For these entrepreneurs, an increase in collateral is matched with an increase in the face
value of debt by the same amount. This has no effect on the probability of default, but
increases the expected profits for the intermediary because more collateral is forfeited.12
For collateral levels between k∗ and w − w1, the probability of default under the optimal
contract xm(k) is strictly between 0 and 1. The intermediary charges a lower face value of
debt to entrepreneurs with more collateral for them to break even, since for any given face
value of debt these entrepreneurs make greater expected payments to the intermediary.
Formally, taking derivatives of B(xm(k), k) = 0 with respect to k, we have:
dxm(k)dk
= − F (xm(k)− k)1− F (xm(k)− k)
< 0.
Using the above expression, we have
dL(xm(k), k)dk
=gf(xm(k)− k)
1− F (xm(k)− k)> 0.
Thus, even though the monopoly lowers the face value of debt for entrepreneurs with
more collateral, the expected loan return L(xm(k), k) increases with collateral, as the
intermediary incurs a smaller monitoring expense due to a lower probability of default.
Finally, for collateral levels above w−w1, the probability of default is zero and the collateral
value does not affect the intermediary’s profit.
3.2. Monopoly deposit rate
To attract funds from lenders, the monopoly intermediary sets a deposit rate. Since
the loan return L to the monopoly intermediary increases with collateral value k under
the loan contracts characterized above, the profit optimization problem of the monopoly
12 Since xm(k) = w∗ + k, we have B(xm(k), k) = w − (p∗ + k), so entrepreneurs’ expected profitsdecrease with the collateral value for k < k∗. However, the expected amount of consumption goods thatthe entrepreneurs receive after completion of the projects, including any returned collateral, is equal tow − p∗, independent of the collateral level. Thus, these entrepreneurs cannot benefit by destroying partof their collateral. The monopoly loan contracts with xm(k) = w∗ + k are no longer incentive compatiblewhen the value of collateralizable assets owned by an entrepreneur is known only to himself, because theentrepreneurs will want to hide part of their assets.
– 13 –
intermediary comes down to choosing the deposit rate r and a “cutoff” collateral value
k, such that only the entrepreneurs with collateral higher than k will be financed. This
problem can be stated as:
maxk,r
∫ k2
k
L(xm(z), z) d(−U(z))− S(r)r
subject to U(k) ≤ S(r). Let rm and km be the solution to the above maximization problem.
An interior solution satisfies the following first-order condition
L(xm(km), km) = rm + S(rm)/S′(rm), (3.2)
and the “market-clearing condition” U(km) = S(rm). The second-order condition is satis-
fied because the loan return increases with the collateral value and by Assumption 2.1 the
marginal cost of loan r + S(r)/S′(r) increases with the deposit rate.13
The first-order condition (3.2) can be interpreted as the familiar monopoly condition
that the marginal revenue equals the marginal cost—the left-hand-side is the marginal
revenue from lending to the entrepreneur with the cutoff value, and the right-hand-side
is the cost of attracting an additional unit of loan from the lenders. Note that unless
the savings function is perfectly elastic, the marginal revenue is strictly greater than the
marginal cost, so that the monopoly intermediary makes a positive profit on the loan to
the marginal entrepreneur.
Due to the ex post monitoring problem not all entrepreneurs get loans; that is, typi-
cally we have km > k1. Entrepreneurs with low collateral values cannot obtain loans, even
though they have access to the same projects as the entrepreneurs who get loans. More
interestingly, the nature of the borrowing constraint that the collateral-poor entrepreneurs
face depends on whether km > k∗ or km < k∗. In the former case, the marginal en-
trepreneur with collateral km earns zero expected profits. The entrepreneurs with lower
collateral values do not get loans simply because they cannot offer the monopoly interme-
diary a loan return that covers the cost of obtaining credit for the monopoly. In contrast,
13 The second-order condition requires that (L(xm(k), k)−r−S(r)/S′(r))U ′(k) have negative derivativeswith respect to k at the solution km, where r is a decreasing function of k via the market-clearing condition.
– 14 –
when km < k∗, the marginal entrepreneur with collateral km earns strictly positive profits.
For the entrepreneurs with collateral values just below km, the monopoly loan contracts
given by (3.1), which are all maximum default loan contracts, would give them positive
expected profits. These entrepreneurs would be willing to give up part of the profits to
avoid being rationed. Due to the ex post monitoring problem, however, there is no cred-
ible way to transfer the profits to the intermediary. As a result, the entrepreneurs with
collateral values just below km do not get loans.
4. Credit Market with Competitive Intermediation
Competitive intermediaries engage in price competitions in both the loan market and the
deposit market. We will argue that this double-sided competition yields a unique “perfect
competitive outcome” where intermediaries earn zero expected profits, with the expected
return on each unit of loan equal to the cost of raising the capital. Competitive loan
contracts and deposit rate will then be characterized. The results will be used in the next
section to compare with monopolistic intermediation.
4.1. Competitive loan contracts
To concentrate on the comparison between monopolistic and competitive intermediation,
we assume that the number of competitive intermediaries is fixed and finite; the exact
number of intermediaries does not affect our results. We model competition among in-
termediaries as the following two-stage game. In the first stage, competing intermediaries
simultaneously offer loan contracts to entrepreneurs and deposit contracts to lenders.14
In the second stage, entrepreneurs and lenders simultaneously decide whether to partici-
pate in the loan market and the deposit market, respectively, and if they do, choose the
best-termed contracts offered by the intermediaries.15
14 In general, intermediaries have incentives to corner one of the two markets in order to securemonopoly power in the other. This is the case when the intermediaries compete in the loan market andin the deposit market sequentially, as in Yanelle (1997).
15 Recall that in Section 2 we have already replaced the decisions of depositors by the aggregate savingsfunction. Here we describe decisions of depositors for the convenience of the presentation.
– 15 –
We focus on symmetric subgame perfect Nash equilibria of the game, consisting of an
identical menu of loan contracts and a common deposit rate offered by each intermediary,
and the decisions by the entrepreneurs and the depositors, such that no agent has an incen-
tive to deviate. We impose two additional restrictions on the strategies of intermediaries,
entrepreneurs and depositors. First, we adopt the convention that on the equilibrium path
when entrepreneurs (depositors) face available contracts that are equally favorable, they
divide their loan applications (savings) evenly among them.16 Each intermediary is dealing
with a continuum of entrepreneurs and can therefore offer risk-free returns to depositors,
as under monopolistic intermediation. In the equilibrium constructed below, this conven-
tion ensures that deposits taken equal loan applications for each intermediary, so that
there is no rationing of either depositors or entrepreneurs. Second, we assume that off the
equilibrium path, the intermediaries may not ration either depositors or entrepreneurs.17
Formally, we say that an intermediary is “bankrupt” if either deposits taken exceed loan
applications or loan applications exceed deposits. In either case, a bankrupt intermediary
receives an arbitrarily large negative payoff.
We claim that a menu of loan contracts and a deposit rate, together with the decisions
by entrepreneurs and depositors satisfying the first restriction above, form a symmetric
subgame perfect equilibrium if (i) each loan contract maximizes the expected profits of the
entrepreneur subject to the intermediary making non-negative profits at the given deposit
rate, and (ii) the total demand for loan contracts equals the total supply of deposits. To
see this, first note that condition (ii) above ensures that there is no rationing of either
depositors or entrepreneurs in equilibrium. Second, for any intermediary to be have a
profitable deviation, a different deposit rate must be offered. A lower rate would not
attract any deposit and would bankrupt the deviating intermediary if loan applications
16 Since it is inefficient for multiple intermediaries to monitor a single entrepreneur, this conventionrequires some coordination among entrepreneurs to ensure that each of them is financed by a singleintermediary.
17 This assumption helps us to focus on the competitive outcome where all capital raised by interme-diaries is loaned out, as in the monopoly case. More importantly, it rules out cornering of markets. Aspointed out by Stahl (1988) and Yanelle (1989), even when intermediaries compete simultaneously in thetwo markets, competitive outcomes can be sensitive to assumptions of the competitive environment. Thesecomplications of competitive intermediation are important, but we ignore them here since our focus is ona direct comparison of monopolistic and competitive intermediation.
– 16 –
x
T-r
B
w1+k w*+k w2+k
Figure 2. B(x, k) and T (x, k)− r as functions of x for fixed k.
are received; a higher rate would attract all deposits, but since the deviating intermediary
may not ration either depositors or entrepreneurs, conditions (i) and (ii) above imply that
there is no a menu of loan contracts for the intermediary to avoid bankruptcy.
We construct an equilibrium under competitive intermediation in two steps. In re-
mainder of this subsection, we fix an arbitrary deposit rate and characterize the menu of
competitive loan contracts that maximize the expected profits of the entrepreneur subject
to the constraint that the intermediary makes non-negative profits. The next subsection we
use this characterization of competitive loan contracts to derive the total demand for loan
contracts as a function of the deposit rate and find the (unique) equilibrium deposit rate so
that the demand equals the supply of deposits. Formally, given any deposit rate r, the loan
contracts for entrepreneurs with collateral k solve the optimization problem of maxx B(x, k)
subject to the participation constraint of the intermediary L(x, k) ≥ r, where B(x, k) is
given by (2.2). The participation constraint can be rewritten as B(x, k) ≤ T (x, k) − r,
where T (x, k) is the sum of the loan return to the intermediary and the expected profits
of the entrepreneur, as given by (2.4). We now show that under this characterization, the
participation constraint of the intermediaries is always binding so that all intermediaries
make zero expected profits.
Figure 2 plots B(x, k) and T (x, k)−r as functions of x, for given r and k, for the case in
which w is uniformly distributed on [w1, w2] and g < w2−w1. The relative position of the
two functions depends on k and r, as we will see in the proof of the following proposition.
– 17 –
Under the uniform assumption, for face value of debt x ∈ [w1 + k, w2 + k], the total net
profit T − r is linear and the expected profit of the entrepreneur B is quadratic. These
special properties do not extend to general output distributions, but two properties shown
in Figure 2 that do hold for any distribution satisfying Assumption 2.2 are: B(x, k) and
T (x, k)− r decrease for x ∈ (w1 + k, w2 + k), and B(x, k) decreases faster than T (x, k)− r
in (w1 +k,w∗+k) and slower in [w∗+k,w2 +k). The latter property is due to Assumption
2.2, because L is the difference between T and B and is increasing in (w1 + k, w∗ + k) and
decreasing in [w∗ + k, w2 + k).
Let xc(k) denote the solution to the problem of maxx B(x, k) subject to B(x, k) ≤T (x, k)−r, as a function of k, for a given r. We now show that the participation constraint
of the intermediaries always binds, and that the competitive loan contract xc(k) is given
by
xc(k)
does not exist, for k < r + k∗ − t∗
satisfies L(xc(k), k) = r, for k ∈ [r + k∗ − t∗, r − w1)
= r, for k ≥ r − w1.
(4.1)
The following proposition gives the result in words, where “competitive intermediation”
refers to simultaneous price competition in the deposit and loan markets by a finite number
of intermediaries which may not ration lenders or borrowers.
Proposition 4.1. Under competitive intermediation, (i) entrepreneurs with low collateral
values do not get loans, (ii) the face value of debt is determined by intermediaries’ zero-
profit condition for intermediate collateral values, (iii) the probability of default is zero for
entrepreneurs with high collateral values.
Proof. Fix r, and consider B(x, k) and T (x, k) − r as functions x for each collateral
value k. Depending on k, the two functions may not intersect, and may intersect more
than once, as in Figure 2. Since B is weakly decreasing for all x, the solution xc(k) to the
problem of maxx B(x, k) subject to B(x, k) ≤ T (x, k) − r corresponds to the intersection
with the smallest x when there are multiple intersections.
For k ≥ r − w1, we have B(0, k) = w ≥ w − r = T (0, k)− r and B(w1 + k, k) = w −(w1+k) ≤ w−r = T (w1+k, k)−r, so B intersects T−r in [0, w1+k]. Since B(x, k) = w−x
and T (x, k) − r = w − r in this range, the intersection is at x = r. The solution xc(k)
– 18 –
in this case is given by the intersection and equals r. For k ∈ [r + k∗ − t∗, r − w1), we
have B(w1 + k, k) > T (w1 + k, k) − r and B(w∗ + k, k) = w − (p∗ + k) ≤ t∗ − r =
T (w∗ + k, k) − r, so the first intersection occurs in (w1 + k, w∗ + k]. The intersection
gives xc(k), which satisfies L(xc(k), k) = r because T = L + B and at the intersection
B = T − r. Finally, for k < r + k∗− t∗, we have B(w∗ + k, k) > T (w∗ + k, k)− r. Because
B(x, k) decreases faster than T (x, k) − r for x ≤ w∗ + k, this means that the two curves
do not intersect for x ≤ w∗ + k. Further, since B(x, k) decreases slower than T (x, k) − r
for x ∈ (w∗ + k, w2 + k), and since both T − r and B become constant for x ≥ w2 + k, it
follows from B(w∗ + k, k) > T (w∗ + k, k) − r that the two functions do not intersect for
x > w∗ + k either. Since B lies above T − r for all x, there is no solution to the problem
of maxx B(x, k) subject to B(x, k) ≤ T (x, k)− r. Q.E.D.
We can explain the intuition behind Proposition 4.1 as follows. In competing for the
loan business, intermediaries strive to maximize the profits of the entrepreneurs subject to
their own participation constraint that the loan return is at least r. The highest return on
the loan to an entrepreneur with collateral k is k+t∗−k∗, achieved by the maximum default
contract with face value w∗ + k. Entrepreneurs with too little collateral (k < r + k∗ − t∗)
do not get loans because it is not feasible for the intermediary to make non-negative
profits from a loan to these entrepreneurs. For all entrepreneurs with adequate collateral
(k ≥ k∗ + r− t∗), since the expected payment P from the entrepreneur decreases with the
face value of debt x, competition among intermediaries means that the face value of debt
will be the lowest that allows the intermediaries to break even. Thus, the participation
constraint of the intermediaries always binds. For entrepreneurs with intermediate values
of collateral (k∗+ r− t∗ ≤ k < r−w1), the participation constraint binds at a point where
the probability of default is strictly between 0 and 1. (This is the case depicted in Figure
2.) Note that for the lowest collateral value k = k∗+ r− t∗ in this range, the two functions
B(x, k) and T (x, k) − r are tangent at w∗ + k, so that the competitive loan contract is
the maximum default contract. For entrepreneurs who can provide sufficient collateral
(k ≥ r − w1), full collateralization is achieved: they pay back r to the intermediaries
regardless of the realized output w, which allows competitive intermediaries to just break
even, and collateral values are so high that default will never occur.
– 19 –
Under the competitive loan contracts described by Proposition 4.1, the expected prof-
its of entrepreneurs increase in collateral value k. To see this, note that for the same loan
contract, entrepreneurs with more collateral are less likely to default. This tends to reduce
the monitoring expenses incurred by the intermediaries and increase the loan returns to
them. Competition among the intermediaries then forces down the face value of debt, and
allows the entrepreneurs to keep more of the project output. Formally, for the intermediate
collateral values k ∈ (k∗ + r − t∗, r − w1), we can differentiate L(xc(k), k) = r to get
dxc(k)dk
=−F (xc(k)− k)− gf(xc(k)− k)
1− F (xc(k)− k)− gf(xc(k)− k). (4.2)
Since xc(k) < w∗ + k, equation (2.5) and Assumption 2.2 imply that the denominator
in the above equation is positive. Thus, face value xc(k) decreases with the collateral
value k, implying that the probability of default decreases and the expected profits to the
entrepreneurs increase.
4.2. Competitive deposit rate
Since the expected profits of entrepreneurs increase in collateral value k, the equilibrium
deposit rate is associated with a cutoff collateral value. Only those entrepreneurs with
collateral values greater than this cutoff value obtain loans in equilibrium. To determine
the equilibrium deposit rate and the cutoff value, we first characterize the cutoff collateral
level as a function of deposit rate r, or the demand function for loans. This is accomplished
by combining Proposition 4.1 with the participation constraint of the entrepreneur. From
Proposition 4.1, the lowest collateral value that allows a competitive intermediary to break
even is k∗ + r − t∗, and the corresponding competitive contract is the maximum default
contract. We distinguish two cases, depending on the deposit rate r.
For r lower than t∗, we argue that the cutoff collateral value is precisely k∗ + r − t∗.
Recall that k∗ is the highest collateral value such that under the maximum default contract
the entrepreneur makes non-negative profits. For deposit rate r lower than t∗, we have
k∗+r−t∗ < k∗, so that the maximum default contract for collateral value k∗+r−t∗ allows
a competitive intermediary to break even and the entrepreneur to make positive profits.
The cutoff collateral value in this case must be k∗ + r− t∗: if it were lower, intermediaries
– 20 –
would not be able to cover the cost of capital (this is shown in Proposition 4.1); if it were
higher intermediaries would deviate and make positive profits by offering the maximum
default contract to the entrepreneurs who do not get loans. Note that for this portion of
the demand function (i.e. r < t∗), the marginal entrepreneurs with the cutoff collateral
value k∗ + r − t∗ make positive profits.
For any deposit rate r higher than t∗, we have k∗ + r − t∗ > k∗. The competitive
contract given by (4.1) for entrepreneurs with collateral value k∗ + r − t∗, which is the
maximum default contract, would leave them with negative profits. In this case, the cutoff
value k is higher than k∗ + r − t∗, and is determined by the entrepreneur’s participation
constraint B(xc(k), k) = 0.
We can now show that the demand function defined above is downward sloping.18
The part corresponding to r < t∗ is downward sloping: if the deposit rate is r, the cutoff
collateral value is k∗ + r − t∗, so a higher deposit rate means a higher cutoff and hence a
lower demand for loans. The part corresponding to r > t∗ is downward sloping too: using
equation (4.2), differentiating B(xc(k), k) = 0 with respect to r and noting that xc(k) also
directly depends on r through L(xc(k), k) = r, we have
dk
dr=
1− F (xc(k)− k)gf(xc(k)− k)
,
which is positive. Thus, a higher deposit rate r means a higher cutoff collateral value k
that satisfies B(xc(k), k) = 0 and hence a lower demand for loans.
The deposit rate rc and the cutoff collateral value kc in the competitive equilibrium can
then be jointly determined by the intersection of the supply function S(r) and the demand
function characterized above. Since both the demand and supply functions are continuous
and monotone, a unique crossing of these two functions exists if supply exceeds demand
at the highest deposit rate and demand exceeds supply at the lowest rate. Sufficient
conditions are easy to find. The highest deposit rate at which there is a positive demand
for loans is w. Recall that the sum of the expected loan return and the entrepreneur’s
18 Note that the demand function is continuous at t∗. For the part corresponding to r ≤ t∗, the cutoffk is equal to k∗ at r = t∗. For the part corresponding to r ≥ t∗, at r = t∗ the entrepreneur’s participationconstraint B(xc(k), k) = 0 is satisfied for k = k∗ and xc(k) = w∗ + k∗, so the cutoff k is also equal to k∗.
– 21 –
profit is at most w, when the probability of default is zero. For both the intermediary and
the entrepreneur to break even when the deposit rate is w, the collateral level must be at
least w − w1 so that default will never occur (see equation 4.1). The demand for loans at
r = w is therefore U(w−w1). The lowest deposit rate is 0, in which case any entrepreneur
can afford a loan, so the demand is U(k1). Then, if S(w) > U(w−w1) and S(0) < U(k1),
a competitive equilibrium exists and is unique.
The equilibrium deposit rate rc may be located on either of the two parts of the de-
mand function. If rc ≥ t∗ (or equivalently kc ≥ k∗), marginal entrepreneurs with collateral
kc make zero profits, while if rc < t∗ (or equivalently kc < k∗), marginal entrepreneurs
earn positive expected profits. In the first case, the marginal entrepreneur with collat-
eral kc earns zero expected profits. The entrepreneurs with lower collateral values do not
get loans because the deposit rate rc is too high: there are no loan contracts which pro-
vide the competitive return r to the intermediaries and make positive profits for these
entrepreneurs. In contrast, when kc < k∗, the marginal entrepreneur with collateral kc
earns strictly positive profits. As in the case of monopolistic intermediation when km < k∗,
there is a discontinuity in the expected profits of the entrepreneurs at the cutoff collat-
eral value kc: the entrepreneurs with collateral values just below kc are not financed even
though the marginal entrepreneur obtains a loan and makes positive expected profits. The
entrepreneurs with collateral just below kc would be willing to accept the loan contract for
the marginal entrepreneur and promise part of the profits to the intermediaries. But the
loan returns are already at their unconstrained maximum for the marginal entrepreneur.
There is no credible way to transfer the profits to the intermediaries due to the ex post
monitoring problem.
5. Does Intermediary Market Structure Matter?
In the previous two sections we have analyzed two representative market structures of inter-
mediation: monopoly and competition. In both cases, capital is allocated to entrepreneurs
with sufficient collateralizable assets. There is a transfer of surplus from the monopoly in-
termediary to entrepreneurs as one shifts from monopolistic intermediation to competitive
– 22 –
intermediation. But besides this transfer of surplus, does intermediary market structure
matter?
5.1. Comparison results
In this section we compare the performance of the credit market under the two structures.
Two variables are important in the comparison. One is the equilibrium deposit rate, since
it determines the total amount of borrowing and lending in both cases. The other one
is the schedule of the face value of debt, since the face value of debt in a loan contract
determines the probability of default and hence the amount of goods produced in excess
of resources consumed by costly monitoring. We will show below that the credit market
performs unambiguously better under competitive intermediation than under monopolistic
intermediation: under competition the deposit rate is higher and the total profits generated
in each unit of loan contract are greater. We show how the monopoly distortions in terms
of less lending and more bankruptcy arise from the monopoly’s market powers with the
lenders and with the borrowers.
First, comparing the extent of savings mobilization under alternative market struc-
tures of intermediation, we have rm ≤ rc and km ≥ kc.
Proposition 5.1. Competitive intermediaries lends less than a monopoly intermediary.
Proof. Consider the marginal entrepreneur with collateral kc under competition. If
rm > rc, then km < kc so this marginal entrepreneur with collateral kc also gets funded
under monopoly. There are two cases. In the first case, kc > k∗ and rc > t∗. Then,
B(xc(kc), kc) = 0. Since kc > k∗ and the entrepreneur with kc gets a loan under monopoly,
Proposition 3.1 implies B(xm(kc), kc) = 0. In the second case, kc ≤ k∗ and rc ≤ t∗.
Then, from Proposition 4.1, xc(kc) = w∗ + kc. From Proposition 3.1, since kc ≤ k∗
and the entrepreneur with kc gets a loan under monopoly, xm(kc) = w∗ + kc. In both
cases, B(xc(kc), kc) = B(xm(kc), kc), implying xc(kc) = xm(kc) and thus L(xc(kc), kc) =
L(xm(kc), kc). Since L(xc(kc), kc) = rc and rm > rc, we have L(xm(kc), kc) < rm, contra-
dicting the optimality of loan contracts under monopoly. Q.E.D.
– 23 –
The intuition behind the above proof is simple. If under monopoly the deposit rate
is higher and more entrepreneurs get loans, then the last entrepreneur to get a loan under
competition is also funded under monopoly. Since competitive intermediaries always make
zero-profits, the monopoly intermediary makes negative profits on the loan to this marginal
entrepreneur due to a higher deposit rate, which is impossible. The above proof formalizes
this argument by showing that this last entrepreneur would be funded with the same loan
contract under monopoly and under competition.
The result that competitive intermediaries lends less than a monopoly intermediary
may not seem surprising. After all, under-production distortions are often associated
with monopolies. However, the monopoly intermediary in our model is not a usual one
because it has monopoly power over both the borrowers and the lenders. Are the monopoly
distortions in mobilizing less savings the result of monopoly power in the loan market or in
the deposit market? The proof of Proposition 5.1 suggests that the distortions arise from
the monopoly power in the deposit market. The basic point is that marginal entrepreneurs
in both cases have the same form of loan contract, and the monopoly lends less because
the cost of capital for the monopoly equals r + S(r)/S′(r), which is higher than the cost
r for competitive intermediaries. The standard result in industrial organization models
that a greater elasticity of S(r) reduces the monopoly power then implies that in our
model of intermediation the monopoly distortions in terms of less lending competitive
intermediation are smaller if the savings function is more elastic.
The next result demonstrates monopoly distortion in each loan contract granted.
Comparing the face value of debt for entrepreneurs with any fixed collateral value k, we
have xm(k) ≥ xc(k).
Proposition 5.2. For entrepreneurs with any fixed collateral value who obtain loans
under alternative market structures, default rates are higher under monopolistic than under
competitive intermediation.
Proof. By Proposition 5.1, any entrepreneur with collateral k > km is also funded
under competition. There are two cases. In the first case, k > k∗. Then, by Proposition
3.1, B(xm(k), k) = 0. In contrast, Proposition 4.1 implies that under competition all en-
trepreneurs that obtain loans earn non-negative expected profits, implying xc(k) ≤ xm(k).
– 24 –
In the second case, k < k∗. Then, by Proposition 3.1, xm(k) = w∗ + k. But under
competition, xm(k) ≤ w∗ + k, with strict inequality except for the marginal entrepreneur
with collateral kc. Thus, xc(k) ≤ xm(k). In either case, default rates are lower under
competition for each k above km. Q.E.D.
The intuition behind Proposition 5.2 can be simply stated as follows. Raising the
face value of debt in the loan contract necessarily increases the expected payment to
the intermediary. Under competition the loan contract minimizes the expected payment
subject to the zero-profit condition of intermediaries, while under monopoly it maximizes
the expected payment subject to the monitoring cost incurred by the intermediary and the
participation of the entrepreneur. Thus, the face value of debt is higher under monopoly
than under competition. Higher face value of debt under monopoly implies higher default
rates for each collateral value.
Comparing Propositions 3.1 and 4.1, we see that the monopoly distortion in terms
of greater default rates for each loan granted varies across entrepreneurs with different
collateral values. The monopoly distortion is the greatest for entrepreneurs with little
collateral. Indeed, for entrepreneurs with enough collateral (greater than w − w1) full
collateralization is achieved and the default rate is zero under both monopolistic and
competitive intermediation. The reasons for the monopoly distortion of more bankruptcy
do not depend on the monopoly’s market power with lenders as represented by the elasticity
of savings function. The proof of Proposition 5.2 makes it clear that even if there is
a competitive investment opportunity for depositors, which would force the monopoly
deposit rate down to the same level as the competitive rate, monopoly debt contracts
would still have higher face value of debt and hence higher default rates than competitive
debt contracts. The monopoly distortion of higher bankruptcy rates arises solely from the
market power with borrowers.
5.2. Comparative statics
Propositions 5.1 and 5.2 can be used to compare impacts on the credit market when ex-
ogenous changes in the economy occur. We are particularly interested in different impacts
on the credit market as available credit or the monitoring cost changes.
– 25 –
First consider what happens when available credit increases as represented by an
outward shift of the savings function S(r). Under both monopolistic intermediation and
competitive intermediation, there will be an expansion of credit to collateral-poor en-
trepreneurs.19 Under monopoly, a lower cutoff collateral value results, with no changes
in the loan contracts for entrepreneurs who obtained loans before the expansion and thus
no change in the default rate (see equation 3.1). Gains in loan businesses and profits are
absorbed by the monopoly intermediary. Under competition, a lower equilibrium deposit
rate due to the credit expansion affects all loan contracts for entrepreneurs (see equa-
tion 4.1). The face value of debt decreases for each loan, and default rates fall across all
entrepreneurs. Thus, while a monopoly intermediary internalizes the impact of a credit
expansion, competitive intermediaries spread the impact throughout the market.
Next, consider the impact of changes in the monitoring cost g. Under both monopo-
listic and competitive intermediation, a reduction in the cost decreases the cutoff collateral
value and increases the deposit rate. The impact on bankruptcy rates, however, is differ-
ent. A smaller monitoring cost makes it profitable for a monopoly intermediary to extend
credit to more entrepreneurs by increasing the deposit rate. However, since the monitoring
expense is directly incurred by the intermediary, a cost reduction has no effect on the face
value of debt, and hence no effect on bankruptcy rates (see equation 3.1). The benefits
of the reduction of the monitoring expense accrue to the monopoly intermediary. In con-
trast, under competitive intermediation, not only will more savings be mobilized because
a decrease in the monitoring cost moves the demand function to the right, but also all
entrepreneurs that obtain loans in equilibrium will benefit from a smaller monitoring cost
(see equation 4.1). Although the monitoring cost is directly incurred by the intermediaries,
competition forces them to transfer the greater surplus to the entrepreneurs through lower
face values of debt.
We summarize the findings in this subsection in the following proposition.
19 The opposite scenario occurs when entrepreneurs with little collateral are squeezed out of the loanmarket first as loanable funds become scarce. This is what Holmstrom and Tirole (1997) call “flightto quality,” and our model provides an alternative explanation. In their model of intermediation, themonitoring problem is ex ante instead of ex post and entrepreneurs’ own capital is liquid but insufficientwithout outside funding. When outside capital becomes scarce in a credit crunch, entrepreneurs withlittle capital of their own lose funding because they have too little at stake and cannot credibly committo proper management of their projects.
– 26 –
Proposition 5.3. In a market expansion due to either an increase in the credit supply or
a decrease in the monitoring cost, the default rate for the pre-existing loans is unaffected
under monopolistic intermediation while it falls under competitive intermediation.
Our results have important policy implications. Credit market liberalization is espe-
cially crucial in economies with an inelastic supply of credit and significant monitoring
costs. Government policies of promoting savings may have limited result in facilitating
growth of the credit market if the intermediary sector remains uncompetitive, because the
benefits of an expansion of credit mostly accrue to the intermediaries with no impact on
bankruptcy rates or the demand for credit by entrepreneurs. Similarly, in a monopolized
credit market reducing monitoring cost merely transfers greater profits to the monopoly
intermediary without reducing bankruptcy rates.
5.3. Robustness analysis
To conclude this section we provide two extensions of our main results regarding the
comparison between monopolistic and competitive intermediation. First, we have assumed
that a borrower’s collateral is worth the same to the lender (or the intermediary) after
liquidation as it is to the borrower. However, one characteristic of collateral may be
that lenders often value liquidated collateral less than the borrower, even if there is no
direct cost in liquidating the collateral. For example, the liquidation value of collateral
to the intermediary can be only a fraction of its original value to the borrower. Imagine
the extreme case that the value of liquidated collateral is 0 to any intermediary. In this
case, we need to make the assumption that the distribution function F is weakly convex,
which replaces Assumption 2.2 and guarantees that the expected gross return L for the
intermediary is concave in the face value x. Then, we can still write the face value of debt
in a maximal default contract as k + w∗, where w∗ is no longer a constant but a function
of k.
The rest of the analysis goes through as before. The immediate effect of introducing
zero liquidation value of collateral is that the expected loan return L(x, k) to the interme-
diary shifts downward for any face value of debt x and collateral level k. Under monopoly,
– 27 –
the resulting effect on loan contracts is a reduced face value of debt and a reduced default
rate. Under competition, the result is exactly the opposite. The different effects of zero
liquidation value arise because the maximal default contract becomes a less effective way
for the monopoly intermediary to exact profits, while recovering the deposit rate promised
to lenders becomes more difficult for competitive intermediaries.
Although introducing zero liquidation value reduces the gap in terms of default rate
between monopoly and competition, our comparison results regarding the two intermediary
market structures (Propositions 5.1and 5.2) remain valid. Relative to competitive inter-
mediation, monopolistic intermediation still leads to two distortions: higher default rates
and less aggregate lending. Introducing zero liquidation value changes both monopoly and
competitive loan contracts, but the face value of debt remains higher under monopoly be-
cause it maximizes the expected payment to the monopoly intermediary while it minimizes
the expected payment to competitive intermediaries. Moreover, introducing zero liquida-
tion value does not alter the fact that marginal entrepreneurs under both monopoly and
competition have the same form of contract (maximal default contract), so the monopoly
lends less due to a higher cost of raising capital from lenders.
For the second extension, imagine that the value of collateral that each entrepreneur is
no longer public information but instead is known only to the entrepreneur. Additionally,
entrepreneurs can choose to put up only part of their collateral in the loan contract with
the intermediary. Since the entrepreneurs who can provide higher levels of collateral make
more profits, competitive intermediation is not affected by the alternative assumption
of private collateral information. The same is true under monopolistic intermediation
when the marginal entrepreneurs make zero expected profits (km > k∗). However, when
the marginal entrepreneurs make positive expected profits (km < k∗), the loan contracts
for entrepreneurs with collateral level below k∗ do not provide the correct incentives for
entrepreneurs to reveal the value of their assets as B(xm(k), k) decreases in k.
We can show that the optimal strategy for the intermediary is to choose a cutoff level of
collateral and give loans only to those who can provide collateral exceeding the cutoff level.
Further, it is optimal for the monopoly to offer the maximum default contract to marginal
entrepreneurs with the cutoff collateral level in order to minimize the surpluses given away.
– 28 –
The face value of debt for any reported collateral level about the cutoff is determined by the
incentive compatibility constraint that all entrepreneurs who obtain loans make at least the
expected profits as the marginal entrepreneur does under the maximum default contract.
The important difference between the case of private information and the case of public
information is that, despite the same concavity of L(x, k), the solution the monopoly’s
optimization problem under private collateral information turns out not to have an interval
of collateral levels where the entrepreneur’s incentive constraint is slack.
Under monopolistic intermediation, whether collateral information is public or pri-
vate does not affect the extent of savings mobilization. This is because the marginal
entrepreneur’s contract takes the same form of the maximum default contract under the
two information structures. Since in either case the first-order condition requires the ex-
pected loan return from the marginal entrepreneur to equal the marginal cost of obtaining
the loan, the cutoff collateral level must be the same. Thus, monopolistic intermediation
with private collateral information involves the same distortion of less lending as com-
pared to competitive intermediation. This is consistent with our conclusion that this type
of distortion arises from monopoly power in the deposit market.
The difference between the two information structures is in bankruptcy rates. The loan
contracts granted by a monopoly intermediary when collateral is private information have
smaller probabilities of default than when collateral information is public. This is because
the monopoly must give incentives, in terms of lower face value of debt, for entrepreneurs
to reveal their true collateral level. Thus, consistent with our observation that monopoly
distortions of more bankruptcy arise from monopoly power in the loan market, a weakening
in monopoly power with borrowers due to private collateral information improves loan
contract performance in terms of bankruptcy rates. Nonetheless, compared to competitive
intermediation, bankruptcy rates are still high.
6. Concluding Remarks
Our paper belongs to the extensive line of literature that compares monopolistic interme-
diation to competitive intermediation. Much of this literature studies the type of frictions
– 29 –
and externalities that can make a concentrated banking sector preferable, in terms of
greater bank lending or smaller default risk, to a perfectly competitive one. Petersen and
Rajan (1995) show that with incomplete markets and imperfect information, monopoly
power can increase the incentive of intermediaries to finance new firms with no record of
performance. Winton (1997) assumes fixed intermediation costs and adoption external-
ities, and argues that the belief of investors that small intermediaries are risky can be
self-fulfilling because of imperfect diversification. Allen and Gale (2000) show that due to
limited liability of the bank and their use of debt contracts, an increase in competition in
the banking sector has the effect of encouraging banks’ risk-seeking behavior.
The present paper differs from those cited above by focusing on the credit risk of
the real sector of the economy, and on the total amount of investment when borrowers
are heterogeneous and loan contracts are collateralized. We offer two contrasting results:
monopolistic intermediation induces less lending and higher bankruptcy rates than com-
petitive intermediation. We are not aware of any empirical works on comparing default
risks under different intermediary market structures, but there is a body of evidence that
is broadly consistent with the result that a high level of concentration reduces credit access
for borrowers with little collateralizable assets.
In a cross-sectional study of U.S. firms, Black and Strahan (2000) find a smaller num-
ber of newly financed firms in states with a higher level of bank concentration. Extending
the international dataset compiled by Rajan and Zingales (1998), Cetorelli and Gambera
(2001) find that bank concentration has on average a depressive effect on industrial growth.
In addition, Berger, Kashyap, and Scalise (1995) establish direct evidence that larger banks
devote smaller proportions of their assets to small business lending. Similarly, Berger and
Udell (1996) find that large banks focus more on financially secure loans, as predicted
by our model. These empirical findings are supported by related studies that examine
the effect of bank M&A on small business lending, including Peek and Rosengren (1998),
Strahan and Weston (1998), Berger, Saunders, Scalise and Udell (1998), and Avery and
Samolyk (2000). The general finding is that post-M&A, large banks reduce lending to
small unsecured businesses. Moreover, Berger, Goldberg and White (2001) find that the
post-M&A reduction in small business lending by large banks has an external effect on
– 30 –
lending by smaller banks. To the extent that a merged bank gains market power, such
empirical evidence provides additional support to the prediction of our model.
Within our framework of costly monitoring, we have made a number of assumptions to
facilitate the analysis. One assumption is that the size of the loan is fixed and independent
of collateral level. When intermediaries choose the size of each loan as well as the face
value of debt, a higher face value of debt may correspond to a smaller size of the loan and
therefore may not increase the default risk. The other assumption is that all entrepreneurs
have the same investment project. When projects differ, there will be a second-degree
price discrimination role for the risky debt contract—the monopoly intermediary will want
to design a schedule of debt contracts so as to induce self-selection by entrepreneurs—
in addition to the role of perfect discrimination according to collateral value studied in
the present paper. Relaxing these and other assumptions remains our immediate target
for future work. Finally, our model focuses on the risk of default and on the role of
collateral, while oversimplifying other aspects of financial intermediation as stressed by
Broecker (1990), Petersen and Rajan (1995), Winton (1997) and Allen and Gale (2000).
How to incorporate project screening, relationship lending, imperfect diversification and
limited bank liability into our framework remains a challenge that needs to be met in
future research.
References
Allen, F., and D. Gale (2000): “Comparing Financial Systems,” MIT Press, Cambridge,MA.
Altman, E. (1984): “A Further Empirical Investigation of the Bankruptcy Cost Question,”Journal of Finance 39, 1067-89.
Avery, R. B., and K. A. Samolyk (2000): “Bank Consolidation and the Provision of Bank-ing Services: The Case of Small Commercial Loans,” Working Paper, Federal DepositInsurance Corporation.
Barro, R. (1976): “The Loan Market, Collateral, and Rates of Interest,” Journal of Money,Credit and Banking 10, 170-83.
Berger, A. N., L. G. Goldberg, and L. J. White (2001): “The Effects of Competition onthe Supply of Small Business Credit,” European Financial Review 5, 115-139.
– 31 –
Berger, A. N., A. K. Kashyap, and J. M. Scalise (1995): “The Transformation of the U.S.Banking Industry: What a Long Strange Trip It’s Been,” Brooking Papers on EconomicActivity 2, 55-218.
Berger, A. N., A. Saunders, J. M. Scalise, and G. F. Udell (1998): “The Effect of BankMergers and Acquisitions on Small Business Lending,” Journal of Financial Economics 50,187-229.
Berger, A. N., and G. F. Udell (1996): “Universal Banking and the Future of SmallBusiness Lending,” in A. Saunders and I. Walter eds., Financial System Design: The Caseof Universal Banking, Irwin, Burr Ridge, IL, 559-627.
Bernake, B. and M. Gertler (1989): “Agency Costs, Net Worth, and Business Fluctua-tions,” American Economic Review 79, 14-31.
Besanko, D. and A. Thakor (1987): “Collateral and Rationing: Sorting Equilibria inMonopolistic and Competitive Credit Markets,” International Economic Review 28, 671-89.
Black, S.E. and P.E. Strahan (2000): “Entrepreneurship and Bank Credit Availability,”Working Paper, Federal Reserve Bank of New York.
Broecker, T. (1990): “Credit-worthiness Tests and Interbank Competition,” Econometrica58, 429-452.
Cetorelli, N. and M. Gambera (2001): “Banking Market Structure, Financial Dependenceand Growth: International Evidence from Industry Data,” Journal of Finance 56, 617-48.
Diamond, D. (1984): “Financial Intermediation and Delegated Monitoring,” Review ofEconomic Studies 51, 393-414.
Franks, J. and W. Torous (1994): “How Firms Fare in Workouts and Chapter 11 Reorga-nization,” Journal of Financial Economics 35, 349-70.
Gale, D. and M. Hellwig (1985): “Incentive-Compatible Debt Contracts: The One PeriodProblem,” Review of Economic Studies 52, 647-63.
Holmstrom, B. and J. Tirole (1997): “Financial Intermediation, Loanable Funds and theReal Sector,” Quarterly Journal of Economics 112, 663-91.
Marquez, R. (2002): “Competition, Adverse Selection, and Information Dispersion in theBanking Industry,” Review of Financial Studies 15, 901-926.
Peek, J., and E. S. Rosengren (1998): “Bank Consolidation and Small Business Lending:It’s not Just Bank Size that Matters,” Journal of Banking and Finance 22, 799-819.
Petersen, M.A. and R. Rajan (1995): “The Effect of Credit Market Competition on Lend-ing Relationship,” Quarterly Journal of Economics 110, 407-43.
– 32 –
Rajan, R. (1992): “Insiders and Outsiders: The Choice between Informed and Arm’s-length Debt,” Journal of Finance 47, 1367-1400.
Rajan, R. and L. Zingales (1998): “Financial Dependence and Growth,” American Eco-nomic Review 88, 559-86.
Riordan, M. (1993), “Competition and Bank Performance: A Theoretical Perspective,” inMayer, C. and X. Vives, editors, Capital Markets and Financial Intermediation, CambridgeUniversity Press.
Schmidt-Mohr, U. (1997): “Rationing versus Collateralization in Competitive and Mo-nopolistic Credit Markets with Asymmetric Information,” European Economic Review 41,1321-42.
Shaffer, S. (1998), “The Winner’s Curse in Banking,” Journal of Financial Intermediation7, 359-392.
Stahl, D.O. (1988): “Bertrand Competition for Input and Walrasian Outcomes,” AmericanEconomic Review, 78, 189-201.
Stiglitz, J. and A. Weiss (1981): “Credit Rationing in Markets with Imperfect Informa-tion,” American Economic Review 71, 393-410.
Strahan, P. E., and J. Weston (1998): “Small Business Lending and the Changing Structureof the Banking Industry,” Journal of Banking and Finance 22, 821-845.
Townsend, R. (1979): “Optimal Contracts and Competitive Markets with Costly StateVerification,” Journal of Economic Theory 21, 265-93.
Williamson, S. (1987a): “Costly Monitoring, Optimal Contracts and Equilibrium CreditRationing,” Quarterly Journal of Economics 102, 135-45.
Williamson, S. (1987b): “Financial Intermediation, Business Failures, and Real BusinessCycles,” Journal of Political Economy 95, 1196-212.
Winton, A. (1997): “Competition among Financial Intermediaries When DiversificationMatters,” Journal of Financial Intermediaries 6, 307-46.
Yanelle, M. (1989): “The Strategic Analysis of Intermediaries,” European Economic Re-view 33, 294-301.
Yanelle, M. (1997): “Banking Competition and Market Efficiency,” Review of EconomicStudies 64(2), 215-39.
Yosha, O. (1997): “Diversification and Competition: Financial Intermediation in a LargeCournot-Walrasian Economy,” Journal of Economic Theory 75, 64-88.
– 33 –